Financial report interpretation｜Baidu's three major businesses are accelerating

This year, as the intelligent cloud business gradually gains self-sustaining capabilities, Baidu has passed through its gear-shifting period. Its current focus is to use generative AI to accelerate transformation.

The year 2024 is considered a key year for the accelerated implementation of large models. Investors are generally concerned about the progress of generative AI (Artificial Intelligence) implementation. Therefore, the first quarter financial reports of various technology companies have attracted much attention. Baidu is one of them.

On May 16, Baidu Group (HKG:9888, NASDAQ:BIDU) released its financial report for the first quarter of 2024. Baidu Group holds controlling interest in iQIYI, and iQIYI's revenue is consolidated within Baidu Group. To truly reflect Baidu's own performance, this article only discusses the performance changes of Baidu Core (excluding iQIYI from Baidu Group).

This quarter, Baidu presented a robust financial report. The report shows that Baidu Core's operating revenue in the first quarter was 23.8 billion yuan, a year-on-year increase of 3.5%. The adjusted net profit attributable to Baidu Core was 6.63 billion yuan, a year-on-year increase of 25.8%, and Baidu Core's operating profit was 4.54 billion yuan, a year-on-year increase of 10.9%. The operating profit margin was 19.1%, an increase of 1.3 percentage points year-on-year.

Investors responded positively to Baidu's performance this quarter. When Baidu released its financial report, the Hong Kong stock market had already closed. However, as of the close on May 16, Eastern Time, Baidu's stock price rose slightly by 1.6%, closing at $112.56, with a total market value of $39.5 billion.

Baidu has three major businesses: online marketing (mobile, search), intelligent cloud, and intelligent driving. All three businesses are directly related to AI. In Baidu's AI transformation, online marketing provides cash flow, intelligent cloud is the current focus of transformation, and intelligent driving is the future focus of transformation.

Online marketing revenue was 17 billion yuan, a year-on-year increase of 3%. This business accounts for 71% of Baidu's total revenue. Baidu's management indicated in the earnings call that Baidu is trying to transform the traditional mobile and search experience with generative AI. Currently, about 11% of Baidu's search result pages are AI-generated results. This provides users with more accurate and direct answers. However, online marketing has not yet achieved commercialization through generative AI.

Intelligent cloud revenue was 4.7 billion yuan, a year-on-year increase of 12%. This business accounts for 20% of Baidu's total revenue. Baidu's management disclosed in the earnings call that revenue from generative AI accounts for 6.9% of the total intelligent cloud revenue. Currently, this part of the revenue mainly comes from model training. With the rapid growth of model inference revenue, the intelligent cloud is expected to usher in a new round of growth.According to information disclosed by Baidu's management in the earnings call for the past five quarters, Baidu's Intelligent Cloud has achieved profitability for five consecutive quarters under non-GAAP (excluding the impact of non-cash factors such as equity incentives and server depreciation, a profit accounting method also adopted by Alibaba Cloud).

Baidu did not disclose progress in its intelligent driving business in the financial report. However, relevant data can be obtained by subtracting the revenue of Intelligent Cloud from the revenue of non-online marketing (including Intelligent Cloud, intelligent driving, Xiaodu speakers, etc.). According to this calculation method, the combined revenue of intelligent driving and Xiaodu speakers is 2.1 billion yuan. This business accounts for 9% of Baidu's total revenue. It still requires continuous investment. Among them, the commercialization of the autonomous taxi service "RoboTaxi" is progressing rapidly, operational efficiency is improving, and it is nearing break-even in some cities.

From 2021 to 2022, Baidu faced the pressure of increasing revenue without increasing profits. It needed the cash flow from its online marketing business to support the cloud and vehicle businesses. As the cloud gradually gained self-sustaining capabilities, Baidu has passed the gear shift period. Its current focus is to use generative AI to accelerate transformation.

Search, AI transformation has just begun

Baidu's online marketing business consists of two parts: search engines and mobile ecosystems. It has long accounted for more than 70% of Baidu's total revenue. The gross margin of the online marketing business is relatively high, usually above 60%. Therefore, it is Baidu's "cash cow." Baidu needs it to remain stable in the long term to create ample development space for Intelligent Cloud and intelligent driving.

The growth rate of Baidu's online marketing business this quarter is relatively flat. Baidu's management explained in the earnings call that the weak macroeconomy has a significant impact. This is because most of Baidu's brand owners come from small and medium-sized enterprises in various industries, which are very sensitive to economic fluctuations. Baidu's management believes that the growth rate of the online marketing business will slow down in the next few quarters, but the fundamentals are relatively stable.

The highlight of the online marketing business is that Baidu is using generative AI to transform its search, mobile products, etc.

A Baidu executive said in August 2023 that Baidu has reconstructed a batch of C-end products with generative AI. After the AI transformation of Baidu Netdisk and Baidu Wenku, user stickiness and membership payment rate have greatly increased. The Wenxin Yiyan App and Baidu Search are also becoming new traffic entrances.

One of Baidu's reconstruction strategies is to enable products such as search, Xiaodu, maps, forums, and car machines to access "intelligent body" capabilities. The "intelligent body" is essentially an AI robot. It can improve the efficiency of user use.Over the past six months, Baidu's strategy of transforming its C-end products with generative AI has achieved certain results. Baidu's management has indicated that the revenue from marketing related to generative AI has increased on a quarter-over-quarter basis, and this trend is expected to continue. This incremental revenue has mitigated the impact of the sluggish macroeconomy. It will also buy time for the transformation of online marketing business.

When the large model craze just started in 2023, a common view in the industry was that generative AI would replace search. It could not bring in advertising revenue for the time being, and might even lead to a decline in search engine revenue. In Google's earnings call, investors have raised similar questions on multiple occasions.

However, the actual situation is that the integration of generative AI and search is becoming increasingly close. Taking Google as an example, the revenue growth rate of its search business has not declined due to generative AI, but has even increased. In the first quarter of 2023, the revenue growth rate of Google's search business was only 2.1%. In the first quarter of 2024, the revenue growth rate of Google's search business has increased to 12.8%.

Baidu's management stated in the earnings call that artificial intelligence is a supplement to traditional search, expanding the potential market. Since the second quarter of 2023, Baidu has been rebuilding Baidu Search with the Wenxin large model. Now, more and more search results are generated by the Wenxin large model. Baidu will continue to use large models to enhance the search experience, which is crucial for increasing the usage rate of search.

In the long run, generative AI will transform the mobile and search user experience, thereby bringing new business increments. There are three directions for this transformation.

Firstly, introducing AI-generated content on the search overview page can provide users with more direct and accurate answers. The search experience will be improved as a result, thereby increasing user stickiness. The earnings call disclosed that about 11% of the search result pages on Baidu are AI-generated results.

Secondly, the marketing system of the search engine can be transformed with generative AI. This can provide brand owners with a series of tools such as creative generation and marketing strategy. Baidu began to provide an AI marketing platform for advertisers in 2023. As a result, the conversion rate of some brand owners' marketing has increased.

Thirdly, providing a brand service center in the search engine. Brand owners can use AI customer service and other tools to attract and convert customers. Traditional search engines often adopt the CPC (Cost Per Click) model. After integrating generative AI into the above model, there is an opportunity to increase the proportion of the CPS (cost per sale, performance-based payment) model. This traffic value is higher and more helpful to enterprise customers.

Therefore, Baidu's founder, chairman, and CEO Robin Li directly stated in this earnings call that search will become one of the "killer" applications of artificial intelligence.

Cloud, has already tasted the AI dividend.Both Chinese cloud service providers and international cloud vendors are generally anticipating that large models will bring new growth opportunities.

The international cloud market experienced a downturn from 2022 to 2023. At that time, influenced by the macroeconomy, enterprises were generally optimizing their IT costs. Amazon Web Services (AWS), Microsoft Azure, and Google Cloud all encountered several quarters of revenue decline. However, by the end of 2023, Amazon AWS, Microsoft Azure, and Google Cloud returned to a normal growth trajectory due to large models.

The situation for Chinese cloud service providers is similar. Starting from 2021, Alibaba Cloud, Tencent Cloud, and Baidu Intelligent Cloud entered a three-year adjustment period. The aforementioned cloud service providers also experienced several consecutive quarters of revenue growth below 10%, and even negative growth.

After the fourth quarter of 2023, Baidu Intelligent Cloud relied on large models to return to a normal growth trajectory—with revenue growth exceeding 10%. In the first quarter of 2024, the revenue growth of Baidu Intelligent Cloud continued to rise to 12%. Although there is still a gap between its revenue growth and that of the top three American cloud service providers, the growth change curve is basically in sync with the top three American cloud service providers. This to some extent indicates that Baidu Intelligent Cloud has seized the initiative in this round of large model technological transformation.

Intelligent cloud has now become Baidu's steadily growing "second curve." It also has a certain self-sustaining ability. According to the information disclosed by Baidu's management in the earnings calls of the past five quarters, Baidu Intelligent Cloud has achieved profitability for five consecutive quarters under the non-GAAP (excluding the impact of non-cash factors such as equity incentives and server amortization; Alibaba Cloud also uses this profit accounting method). Baidu's management stated in the earnings call that, in the long run, they expect the profit margins of AI and large model-related businesses to further improve and be higher than traditional cloud services.

The growth of intelligent cloud is mainly benefiting from the current changes in the computing power structure. Both training and inference of large models require AI chips. Therefore, the growth rate of intelligent computing power (such as GPU and other AI computing power) is far exceeding that of general computing power (CPU computing power). International market research firm IDC estimated in December 2023 that from 2022 to 2027, the annual compound growth rate of China's intelligent computing power scale will reach 33.9%, while the annual compound growth rate of general computing power scale will be 16.6%.

Several Chinese cloud service provider executives have recently expressed a similar view—that since 2023, the structural change in the overall computing power is that the growth rate of intelligent computing power is greater than that of general computing power. However, after 2024, the structural change in intelligent computing power is that the growth rate of inference computing power will be greater than that of training computing power. The reason is that as various industries implement large models in their business operations, a batch of AI-native applications will be rapidly born.

In this earnings call, Baidu's management's judgment on the development of the intelligent cloud business is similar.Firstly, cloud computing is transitioning from general computing power to intelligent computing power, which will reshape the competitive landscape of the cloud market. Baidu has the opportunity to become a leader in the industry.

Secondly, the current growth momentum of intelligent cloud primarily relies on model training. In the near future, the demand for model inference is expected to explode, ushering in a new round of growth opportunities for intelligent cloud.

Thirdly, Baidu's GPU computing power has driven the cross-selling of CPU computing power. Customers have a high recognition of Baidu's GPU computing power, and as a result, they are migrating their CPU computing power needs to Baidu.

Baidu Intelligent Cloud's current revenue scale still lags behind other top cloud vendors. It must continue to maintain rapid growth to catch up with the leaders. It must seize other growth opportunities, such as preparing for the potential explosion of AI applications and the growth of model inference in the future.

At present, Baidu has already taken some actions. Baidu has launched the Qianfan AppBuilder application development platform. Enterprises can use it to develop lightweight, fragmented, and atomic generative AI applications. Baidu is also exploring generative AI applications with a group of software manufacturers and has established an AI application store.

Baidu's management disclosed in the earnings call that in April 2024, the Wenxin large model processed approximately 200 million API (Application Programming Interface) calls per day, a significant increase from the approximately 50 million calls in December 2023. This growth indicates that the adoption rate of the Wenxin large model is continuously increasing, and there will be a huge revenue potential for future model inference.

The continuous development of the cloud requires computing power reserves. Large models, whether for training or inference, require ample computing resources. In 2023, global technology companies with cloud businesses were engaged in an arms race for computing power, purchasing a large number of high-performance GPU servers. Baidu is no exception.

The scale of computing power procurement is usually reflected in capital expenditures. The capital expenditures of technology companies typically include the procurement of servers, the construction of data centers, land leasing, etc. The amortization period for enterprise server procurement is generally 5 to 6 years. The actual procurement expenditure for the current quarter will not be fully recorded in the books but will be amortized over the next 5 to 6 years of capital expenditures.

Baidu needs to be well prepared for the computing power arms race. The core capital expenditures of Baidu can show relevant signs. Starting from the second quarter of 2023, Baidu's core capital expenditures began to grow significantly. The average growth rate over the past four quarters was as high as 61.4%. This increase is not lower than that of Microsoft. Microsoft is one of the more aggressive American cloud vendors in terms of capital expenditures in 2023, with an average quarterly capital expenditure increase of 47%.Cars, Accelerated Commercialization and Implementation

Baidu's intelligent driving business currently consists of three main areas. Autonomous driving, which is the RoboTaxi service under the brand "RoboSense"; intelligent vehicles, primarily providing assisted driving solutions to car manufacturers; and lifestyle travel, building an ecosystem service around Baidu Maps.

Baidu's core strategy is to develop autonomous taxi services and assisted driving businesses based on AI technology, and then expand into the entire travel service system. Among them, autonomous taxis and assisted driving complement each other, serving different user needs. It can also collect high-quality data during operation, which is then fed back to the large autonomous driving model to enhance the maturity of autonomous driving technology.

RoboSense is considered a part of Baidu's autonomous driving business that is progressing well in commercialization and has considerable market potential. Recent data from Huafu Securities shows that the operating cost per kilometer for RoboSense is only 0.81 yuan, saving 58% compared to traditional fuel vehicles and 43% compared to traditional electric vehicles.

Baidu's financial reports usually do not disclose revenue data for its intelligent driving business. This financial report only disclosed operational data for RoboSense. As of April 19, RoboSense has provided more than 6 million rides. In the first quarter, it provided about 826,000 rides, a year-on-year increase of 25%. This number is slightly lower than the 839,000 orders in the fourth quarter of last year. There is a noticeable seasonal fluctuation in the number of travel service orders. Therefore, this data is relatively stable.

Since 2023, due to the maturity of the industry chain and the gradual relaxation of policies, RoboSense has entered a fast track of development. Currently, it has been opened in cities such as Beijing, Shanghai, Guangzhou, Wuhan, Shenzhen, Chongqing, Changsha, and Yangquan, and is available for public trial operation on a regular basis. As of April 19, 2024, RoboSense has provided more than 6 million orders of autonomous driving travel services to the public. In the first quarter of 2024, RoboSense provided about 826,000 rides to the public, a year-on-year increase of 25%.

For RoboSense, the more realistic issue now is commercialization. Therefore, the focus of RoboSense's business from 2023 to 2024 is to improve the regional unit economic benefits in key cities. In this regard, Baidu has adopted two strategies.

First, reduce the cost of the entire vehicle and operating costs. On May 15, Baidu just released the new generation of RoboSense unmanned vehicles, with the cost of the entire vehicle reduced by 60% compared to the fifth generation. The price has dropped to 200,000 yuan.

Second, expand the service coverage and improve operational efficiency. The scale effect of unmanned taxis is obvious. Baidu's management introduced in the financial report conference call that RoboSense's operational coverage in Wuhan has expanded eight times compared to a year ago, currently covering more than 7 million people locally. The market scale has reached 1% of the local ride-hailing market. The operating time of RoboSense has also gradually expanded from the initial off-peak hours to include peak hours, and was extended to 7x24 hours in March this year. The average daily passenger load and the distance per ride for each vehicle are increasing.

The above strategies have enabled RoboSense's profitability to be improving. Baidu's management stated in the financial report conference call that the company plans to expand the scale of the fully unmanned fleet operating in Wuhan to 1,000 vehicles by the end of 2024, more than triple the size at the end of last year. The current focus is still to improve the unit economic model and reduce RoboSense's losses. With improved operational efficiency and reduced costs, RoboSense will first achieve a break-even in Wuhan. Once achieved, it can quickly expand the scale of operation.Baidu's objective plan is for its autonomous driving platform, RoboTaxi, to achieve a break-even point in Wuhan by the end of 2024, with a full profit expected by 2025. If everything proceeds as planned, RoboTaxi could become one of the first platforms globally to realize commercial profitability in autonomous driving.

A recent development in the autonomous driving industry is the accelerated entry of Tesla's Full Self-Driving (FSD) into China. It is widely believed that the introduction of Tesla's FSD could trigger a "catfish effect." This might inspire local companies to accelerate the research and development of autonomous driving technologies and promote the industry towards higher levels of autonomous driving. At that time, Baidu's RoboTaxi might also be one of the beneficiaries.

The continuous and healthy development of all of Baidu's businesses is inseparable from its AI foundation. After the financial report release on May 16, Baidu's founder, chairman, and CEO, Robin Li, stated that Baidu's next step is to make the Wenxin large model series more affordable and efficient. "This will bring more opportunities for Baidu."